netem: bug fixes
Subsystems (4)
Stephen Hemminger
maintainerv8 cover netem: bug fixes
These bugs were found when doing AI-assisted review of sch_netem.c during investigation of the packet duplication recursion problem addressed in Jamal's series.
The fixes cover:
- probability gaps in the 4-state Markov loss model
- queue limit not accounting for reordered packets
- PRNG reseeded on every tc change, breaking reproducibility
- slot configuration not validated (inverted ranges, negative delays, negative limits)
- slot delay arithmetic overflow for ranges above ~2.1 seconds
- negative latency and jitter wrapping to huge time_to_send values via u64 arithmetic
v8 - added check for negative TCA_NETEM_LATENCY64 and TCA_NETEM_JITTER64
- extended slot validation to cover dist_delay, dist_jitter, max_packets and max_bytes
Stephen Hemminger (6):
net/sched: netem: fix probability gaps in 4-state loss model net/sched: netem: fix queue limit check to include reordered packets net/sched: netem: only reseed PRNG when seed is explicitly provided net/sched: netem: validate slot configuration
net/sched: netem: fix slot delay calculation overflow net/sched: netem: check for negative latency and jitter
net/sched/sch_netem.c | 76 ++++++++++++++++++++++++++++++++++++------- 1 file changed, 64 insertions(+), 12 deletions(-)
Signature · 1 line
2.53.0
Stephen Hemminger
maintainerPATCH v8 1/6 net/sched: netem: fix probability gaps in 4-state loss model
The 4-state Markov chain in loss_4state() has gaps at the boundaries between transition probability ranges. The comparisons use:
if (rnd < a4) else if (a4 < rnd && rnd < a1 + a4)
When rnd equals a boundary value exactly, neither branch matches and no state transition occurs. The redundant lower-bound check (a4 < rnd) is already implied by being in the else branch.
Remove the unnecessary lower-bound comparisons so the ranges are contiguous and every random value produces a transition, matching the GI (General and Intuitive) loss model specification.
This bug goes back to original implementation of this model.
--- net/sched/sch_netem.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
net/sched/sch_netem.c +4 -4
@@ -227,10 +227,10 @@ static bool loss_4state(struct netem_sched_data *q)if (rnd < clg->a4) { clg->state =LOST_IN_GAP_PERIOD ;return true; - }else if (clg->a4 < rnd && rnd < clg->a1 + clg->a4) { + }else if (rnd < clg->a1 + clg->a4) { clg->state =LOST_IN_BURST_PERIOD ;return true; - }else if (clg->a1 + clg->a4 < rnd) { + }else { clg->state =TX_IN_GAP_PERIOD ; }
@@ -247,9 +247,9 @@ static bool loss_4state(struct netem_sched_data *q)case LOST_IN_BURST_PERIOD :if (rnd < clg->a3) clg->state =TX_IN_BURST_PERIOD ; -else if (clg->a3 < rnd && rnd < clg->a2 + clg->a3) { +else if (rnd < clg->a2 + clg->a3) { clg->state =TX_IN_GAP_PERIOD ; - }else if (clg->a2 + clg->a3 < rnd) { + }else { clg->state =LOST_IN_BURST_PERIOD ;return true; }
Signature · 1 line
2.53.0
Stephen Hemminger
maintainerPATCH v8 2/6 net/sched: netem: fix queue limit check to include reordered packets
In reply to Stephen HemmingerThe queue limit check in netem_enqueue() uses q->t_len which only counts packets in the internal tfifo. Packets placed in sch->q by the reorder path (__qdisc_enqueue_head) are not counted, allowing the total queue occupancy to exceed sch->limit under reordering.
Include sch->q.qlen in the limit check.
--- net/sched/sch_netem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
net/sched/sch_netem.c +1 -1
@@ -524,7 +524,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,1 <<get_random_u32_below (8 ); } -if (unlikely (q->t_len >= sch->limit)) { +if (unlikely (sch->q.qlen >= sch->limit)) {/* re-link segs, so that qdisc_drop_all() frees them all */ skb->next = segs;qdisc_drop_all (skb, sch, to_free);
Signature · 1 line
2.53.0
Simon Horman
reviewerRe: [PATCH net v8 2/6] net/sched: netem: fix queue limit check to include reordered packets
Quoted 10 lines · Stephen Hemminger
The queue limit check in
netem_enqueue()usesq->t_lenwhich only counts packets in the internal tfifo. Packets placed insch->qby the reorder path (__qdisc_enqueue_head) are not counted, allowing the total queue occupancy to exceedsch->limitunder reordering.Include
sch->q.qlenin the limit check.Fixes f8d4bc455047 ("net/sched: netem: account for backlog updates from child qdisc")Signed-off-by Stephen Hemminger <[email protected]>Reviewed-by Simon Horman <[email protected]>
I acknowledge that Sashiko has provided review of this patch.
In the case of the commentary on the use of sch->q.qlen in place of q->t_len:
this is exactly the intention of this patch; to address shortcomings in commit f8d4bc455047. Follow-up issues can be treated as such.
In the case of feedback on us of qdisc_drop_all(): if this is a problem then it predates this patch and can be addressed separately to this patch-set.
...
Stephen Hemminger
maintainerPATCH v8 3/6 net/sched: netem: only reseed PRNG when seed is explicitly provided
In reply to Stephen Hemmingernetem_change() unconditionally reseeds the PRNG on every tc change command. If TCA_NETEM_PRNG_SEED is not specified, a new random seed is generated, destroying reproducibility for users who set a deterministic seed on a previous change.
Move the initial random seed generation to netem_init() and only reseed in netem_change() when TCA_NETEM_PRNG_SEED is explicitly provided by the user.
--- net/sched/sch_netem.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
net/sched/sch_netem.c +6 -4
@@ -1112,11 +1112,10 @@ static int netem_change(struct Qdisc *sch, struct nlattr *opt,/* capping jitter to the range acceptable by tabledist() */ q->jitter =min_t (s64,abs (q->jitter),INT_MAX ); -if (tb[TCA_NETEM_PRNG_SEED ]) +if (tb[TCA_NETEM_PRNG_SEED ]) { q->prng.seed =nla_get_u64 (tb[TCA_NETEM_PRNG_SEED ]); -else - q->prng.seed =get_random_u64 (); -prandom_seed_state (&q->prng.prng_state, q->prng.seed); +prandom_seed_state (&q->prng.prng_state, q->prng.seed); + } unlock:sch_tree_unlock (sch);
@@ -1139,6 +1138,9 @@ static int netem_init(struct Qdisc *sch, struct nlattr *opt,return -EINVAL ; q->loss_model =CLG_RANDOM ; + q->prng.seed =get_random_u64 (); +prandom_seed_state (&q->prng.prng_state, q->prng.seed); + ret =netem_change (sch, opt, extack);if (ret)pr_info ("netem: change failed\n" );
Signature · 1 line
2.53.0
Stephen Hemminger
maintainerPATCH v8 4/6 net/sched: netem: validate slot configuration
In reply to Stephen HemmingerReject slot configurations that have no defensible meaning:
- negative min_delay or max_delay
- min_delay greater than max_delay
- negative dist_delay or dist_jitter
- negative max_packets or max_bytes
Negative or out-of-order delays underflow in get_slot_next(), producing garbage intervals. Negative limits trip the per-slot accounting (packets_left/bytes_left <= 0) on the first packet of every slot, defeating the rate-limiting half of the slot feature.
Note that dist_jitter has been silently coerced to its absolute value by get_slot() since the feature was introduced; rejecting negatives here converts that silent coercion into -EINVAL. The abs() can be removed in a follow-up.
--- net/sched/sch_netem.c | 29 +++++++++++++++++++++++++++++ 1 file changed, 29 insertions(+)
net/sched/sch_netem.c +29 -0
@@ -827,6 +827,29 @@ static int get_dist_table(struct disttable **tbl, const struct nlattr *attr)return 0 ; } +static int validate_slot (const struct nlattr *attr,struct netlink_ext_ack *extack) +{ +const struct tc_netem_slot *c =nla_data (attr); + +if (c->min_delay <0 || c->max_delay <0 ) { +NL_SET_ERR_MSG_ATTR (extack, attr,"negative slot delay" ); +return -EINVAL ; + } +if (c->min_delay > c->max_delay) { +NL_SET_ERR_MSG_ATTR (extack, attr,"slot min delay greater than max delay" ); +return -EINVAL ; + } +if (c->dist_delay <0 || c->dist_jitter <0 ) { +NL_SET_ERR_MSG_ATTR (extack, attr,"negative dist delay" ); +return -EINVAL ; + } +if (c->max_packets <0 || c->max_bytes <0 ) { +NL_SET_ERR_MSG_ATTR (extack, attr,"negative slot limit" ); +return -EINVAL ; + } +return 0 ; +} +static void get_slot (struct netem_sched_data *q,const struct nlattr *attr) {const struct tc_netem_slot *c =nla_data (attr);
@@ -1040,6 +1063,12 @@ static int netem_change(struct Qdisc *sch, struct nlattr *opt, goto table_free; } +if (tb[TCA_NETEM_SLOT ]) { + ret =validate_slot (tb[TCA_NETEM_SLOT ], extack); +if (ret) + goto table_free; + } +sch_tree_lock (sch);/* backup q->clg and q->loss_model */ old_clg = q->clg;
Signature · 1 line
2.53.0
Re: [PATCH net v8 4/6] net/sched: netem: validate slot configuration
Quoted 12 lines · Stephen Hemminger
Reject slot configurations that have no defensible meaning:
- negative
min_delayormax_delay
-min_delaygreater thanmax_delay
- negativedist_delayordist_jitter
- negativemax_packetsormax_bytesNegative or out-of-order delays underflow in
get_slot_next(), producing garbage intervals. Negative limits trip the per-slot accounting (packets_left/bytes_left <= 0) on the first packet of every slot, defeating the rate-limiting half of the slot feature.
Quoted 7 lines
Note that
dist_jitterhas been silently coerced to its absolute value byget_slot()since the feature was introduced; rejecting negatives here converts that silent coercion into-EINVAL. Theabs()can be removed in a follow-up.Fixes 836af83b54e3 ("netem: support delivering packets in delayed time slots")Signed-off-by Stephen Hemminger <[email protected]>
I acknowledge that Sashiko has provided feedback on this patch.
1. "Does rejecting negative dist_jitter values with -EINVAL cause a regression in userspace ABI backward compatibility? Since the kernel previously accepted these values and silently coerced them using abs(), existing userspace tools or scripts that happen to pass negative values might start failing to configure the qdisc."
This is intended and explicitly explained in the cover letter.
2. "Could directly dereferencing 64-bit fields from the netlink attribute payload cause undefined behavior on strict-alignment architectures? Netlink attribute payloads are typically only guaranteed to be 4-byte aligned because NLA_ALIGNTO is 4, but the __s64 fields within tc_netem_slot like min_delay require 8-byte natural alignment. Performing an 8-byte read from a potentially 4-byte aligned address might cause an alignment fault on certain architectures."
This patch does not change the presence of this problem; and I suspect it is not a problem at all.
Paolo Abeni
maintainerRe: [PATCH net v8 4/6] net/sched: netem: validate slot configuration
Quoted 25 lines · Simon Horman
Quoted 12 lines · Stephen Hemminger
Reject slot configurations that have no defensible meaning:
- negative
min_delayormax_delay
-min_delaygreater thanmax_delay
- negativedist_delayordist_jitter
- negativemax_packetsormax_bytesNegative or out-of-order delays underflow in
get_slot_next(), producing garbage intervals. Negative limits trip the per-slot accounting (packets_left/bytes_left <= 0) on the first packet of every slot, defeating the rate-limiting half of the slot feature.Quoted 7 lines
Note that
dist_jitterhas been silently coerced to its absolute value byget_slot()since the feature was introduced; rejecting negatives here converts that silent coercion into-EINVAL. Theabs()can be removed in a follow-up.Fixes 836af83b54e3 ("netem: support delivering packets in delayed time slots")Signed-off-by Stephen Hemminger <[email protected]>Reviewed-by Simon Horman <[email protected]>I acknowledge that Sashiko has provided feedback on this patch.
Quoted 7 lines
1. "Does rejecting negative
dist_jittervalues with-EINVALcause a regression in userspace ABI backward compatibility? Since the kernel previously accepted these values and silently coerced them usingabs(), existing userspace tools or scripts that happen to pass negative values might start failing to configure the qdisc."This is intended and explicitly explained in the cover letter.
Jamal, given the uAPI implication, could you please double check that the change is fine?
Thanks,
Paolo
Jamal Hadi Salim
maintainerRe: [PATCH net v8 4/6] net/sched: netem: validate slot configuration
Quoted 33 lines · Paolo Abeni
Quoted 31 lines · Simon Horman
Quoted 12 lines · Stephen Hemminger
Reject slot configurations that have no defensible meaning:
- negative
min_delayormax_delay
-min_delaygreater thanmax_delay
- negativedist_delayordist_jitter
- negativemax_packetsormax_bytesNegative or out-of-order delays underflow in
get_slot_next(), producing garbage intervals. Negative limits trip the per-slot accounting (packets_left/bytes_left <= 0) on the first packet of every slot, defeating the rate-limiting half of the slot feature.Quoted 7 lines
Note that
dist_jitterhas been silently coerced to its absolute value byget_slot()since the feature was introduced; rejecting negatives here converts that silent coercion into-EINVAL. Theabs()can be removed in a follow-up.Fixes 836af83b54e3 ("netem: support delivering packets in delayed time slots")Signed-off-by Stephen Hemminger <[email protected]>Reviewed-by Simon Horman <[email protected]>I acknowledge that Sashiko has provided feedback on this patch.
1. "Does rejecting negative
dist_jittervalues with-EINVALcause a regression in userspace ABI backward compatibility? Since the kernel previously accepted these values and silently coerced them usingabs(), existing userspace tools or scripts that happen to pass negative values might start failing to configure the qdisc."
Quoted 4 lines
Quoted 1 line
This is intended and explicitly explained in the cover letter.
Jamal, given the uAPI implication, could you please double check that the change is fine?
It should be fine; at least iproute2 will never allow the kernel to receive a negative number.
Stephen brought up the fact that strtod() could return a -ve number (but at least iproute2 makes sure negative numbers are not carried forward to the kernel).
cheers,
jamal
Quoted 4 lines
Thanks,
Paolo
Stephen Hemminger
maintainerRe: [PATCH net v8 4/6] net/sched: netem: validate slot configuration
Quoted 5 lines · On Thu, 23 Apr 2026 17:12:15 -0400 Jamal Hadi Salim <[email protected]> wrote:
Quoted 4 lines
Quoted 1 line
This is intended and explicitly explained in the cover letter.
Jamal, given the uAPI implication, could you please double check that the change is fine?
Quoted 8 lines
It should be fine; at least iproute2 will never allow the kernel to receive a negative number.
Stephen brought up the fact thatstrtod()could return a -ve number (but at least iproute2 makes sure negative numbers are not carried forward to the kernel).cheers,
jamal
Iproute2 blocks negative values kind of by accident.
The NEXT_IS_NUMBER() macro looks for digit at start of arg. To hit this you need to either use raw netlink or change NEXT_IS_NUMBER() to NEXT_IS_SIGNED_NUMBER() where slot values are parsed.
Jamal Hadi Salim
maintainerRe: [PATCH net v8 4/6] net/sched: netem: validate slot configuration
() ncalls s
Quoted 15 lines · Stephen Hemminger
Quoted 11 lines · On Thu, 23 Apr 2026 17:12:15 -0400 Jamal Hadi Salim <[email protected]> wrote:
Quoted 4 lines
Quoted 1 line
This is intended and explicitly explained in the cover letter.
Jamal, given the uAPI implication, could you please double check that the change is fine?
It should be fine; at least iproute2 will never allow the kernel to receive a negative number.
Stephen brought up the fact thatstrtod()could return a -ve number (but at least iproute2 makes sure negative numbers are not carried forward to the kernel).
Quoted 7 lines
Quoted 2 lines
cheers,
jamalIproute2 blocks negative values kind of by accident.
TheNEXT_IS_NUMBER()macro looks for digit at start of arg. To hit this you need to either use raw netlink or changeNEXT_IS_NUMBER()toNEXT_IS_SIGNED_NUMBER()where slot values are parsed.
For this specific attribute it was intentional:
...
if (get_time64(&slot.dist_jitter, *argv)) { //get_time64 calls strtod() which could set dist_jitter -ve
explain1("slot jitter");
return -1;
}
if (slot.dist_jitter <= 0) { // we reject -ve values
fprintf(stderr, "Non-positive jitter\n");
return -1;
}cheers,
jamal
Stephen Hemminger
maintainerRe: [PATCH net v8 4/6] net/sched: netem: validate slot configuration
Quoted 40 lines · On Fri, 24 Apr 2026 16:17:13 -0400 Jamal Hadi Salim <[email protected]> wrote:
() ncalls s
Quoted 15 lines · Stephen Hemminger
Quoted 11 lines · On Thu, 23 Apr 2026 17:12:15 -0400 Jamal Hadi Salim <[email protected]> wrote:
Quoted 4 lines
Quoted 1 line
This is intended and explicitly explained in the cover letter.
Jamal, given the uAPI implication, could you please double check that the change is fine?
It should be fine; at least iproute2 will never allow the kernel to receive a negative number.
Stephen brought up the fact thatstrtod()could return a -ve number (but at least iproute2 makes sure negative numbers are not carried forward to the kernel).Quoted 7 lines
Quoted 2 lines
cheers,
jamalIproute2 blocks negative values kind of by accident.
TheNEXT_IS_NUMBER()macro looks for digit at start of arg. To hit this you need to either use raw netlink or changeNEXT_IS_NUMBER()toNEXT_IS_SIGNED_NUMBER()where slot values are parsed.For this specific attribute it was intentional:
...
if (get_time64(&slot.dist_jitter, *argv)) { //get_time64 callsstrtod()which could setdist_jitter-veexplain1("slot jitter"); return -1; } if (slot.dist_jitter <= 0) { // we reject -ve values fprintf(stderr, "Non-positive jitter\n"); return -1; }
Quoted 2 lines
cheers,
jamal
I see no cases where negative time value is a valid input. Doing audit of netem:
- slot times
parsing rejects -10ms for min delay and -100ms for max delay
- delay
negative value is accepted, but internally acts like 0
- jitter
negative value is rejected (i.e What is "-2s")
- slot distribution delay
same as regular delay, accepted but internally acts like 0 negative jitter is rejected "Non positive jitter"
Gate:
Codel:
- setting negative target, interval or ce_threshold gives huge value
4.29e+03s
Pi:
- setting negative target gets absolute value
Hhf:
- setting negative value gets 4.29e+03s
So I can't see any valid use for negative intervals
Stephen Hemminger
maintainerPATCH v8 5/6 net/sched: netem: fix slot delay calculation overflow
In reply to Stephen Hemmingerget_slot_next() computes a random delay between min_delay and max_delay using:
get_random_u32() * (max_delay - min_delay) >> 32
This overflows signed 64-bit arithmetic when the delay range exceeds approximately 2.1 seconds (2^31 nanoseconds), producing a negative result that effectively disables slot-based pacing. This is a realistic configuration for WAN emulation (e.g., slot 1s 5s).
Use mul_u64_u32_shr() which handles the widening multiply without overflow.
--- net/sched/sch_netem.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
net/sched/sch_netem.c +2 -3
@@ -659,9 +659,8 @@ static void get_slot_next(struct netem_sched_data *q, u64 now)if (!q->slot_dist) next_delay = q->slot_config.min_delay + - (get_random_u32 () * - (q->slot_config.max_delay - - q->slot_config.min_delay) >>32 ); +mul_u64_u32_shr (q->slot_config.max_delay - q->slot_config.min_delay, +get_random_u32 (),32 );else next_delay = tabledist(q->slot_config.dist_delay, (s32)(q->slot_config.dist_jitter),
Signature · 1 line
2.53.0
Stephen Hemminger
maintainerPATCH v8 6/6 net/sched: netem: check for negative latency and jitter
In reply to Stephen HemmingerReject requests with negative latency or jitter.
A negative value added to current timestamp (u64) wraps
to an enormous time_to_send, disabling dequeue.
The original UAPI used u32 for these values; the conversion to 64-bit time values via TCA_NETEM_LATENCY64 and TCA_NETEM_JITTER64 allowed signed values to reach the kernel without validation.
Jitter is already silently clamped by an abs() in netem_change(); that abs() can be removed in a follow-up once this rejection is in place.
--- net/sched/sch_netem.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
net/sched/sch_netem.c +22 -0
@@ -826,6 +826,16 @@ static int get_dist_table(struct disttable **tbl, const struct nlattr *attr)return 0 ; } +static int validate_time (const struct nlattr *attr,const char *name, +struct netlink_ext_ack *extack) +{ +if (nla_get_s64 (attr) <0 ) { +NL_SET_ERR_MSG_ATTR_FMT (extack, attr,"negative %s" , name); +return -EINVAL ; + } +return 0 ; +} +static int validate_slot (const struct nlattr *attr,struct netlink_ext_ack *extack) {const struct tc_netem_slot *c =nla_data (attr);
@@ -1068,6 +1078,18 @@ static int netem_change(struct Qdisc *sch, struct nlattr *opt, goto table_free; } +if (tb[TCA_NETEM_LATENCY64 ]) { + ret =validate_time (tb[TCA_NETEM_LATENCY64 ],"latency" , extack); +if (ret) + goto table_free; + } + +if (tb[TCA_NETEM_JITTER64 ]) { + ret =validate_time (tb[TCA_NETEM_JITTER64 ],"jitter" , extack); +if (ret) + goto table_free; + } +sch_tree_lock (sch);/* backup q->clg and q->loss_model */ old_clg = q->clg;
Signature · 1 line
2.53.0
Re: [PATCH net v8 6/6] net/sched: netem: check for negative latency and jitter
Quoted 7 lines · Stephen Hemminger
Reject requests with negative latency or jitter.
A negative value added to current timestamp (u64) wraps
to an enormoustime_to_send, disabling dequeue.
The original UAPI used u32 for these values; the conversion to 64-bit time values viaTCA_NETEM_LATENCY64andTCA_NETEM_JITTER64allowed signed values to reach the kernel without validation.
Quoted 6 lines
Jitter is already silently clamped by an
abs()innetem_change(); thatabs()can be removed in a follow-up once this rejection is in place.Fixes 99803171ef04 ("netem: add uapi to express delay and jitter in nanoseconds")Signed-off-by Stephen Hemminger <[email protected]>
This series was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This series was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:
Quoted 15 lines · On Fri, 17 Apr 2026 20:19:38 -0700 you wrote:
These bugs were found when doing AI-assisted review of
sch_netem.cduring investigation of the packet duplication recursion problem addressed in Jamal's series.The fixes cover:
- probability gaps in the 4-state Markov loss model
- queue limit not accounting for reordered packets
- PRNG reseeded on every tc change, breaking reproducibility
- slot configuration not validated (inverted ranges, negative delays, negative limits)
- slot delay arithmetic overflow for ranges above ~2.1 seconds
- negative latency and jitter wrapping to hugetime_to_sendvalues via u64 arithmetic
Quoted 1 line
[...]
Here is the summary with links:
- [net,v8,1/6] net/sched: netem: fix probability gaps in 4-state loss model
https://git.kernel.org/netdev/net/c/732b463449fd - [net,v8,2/6] net/sched: netem: fix queue limit check to include reordered packets https://git.kernel.org/netdev/net/c/4185701fcce6 - [net,v8,3/6] net/sched: netem: only reseed PRNG when seed is explicitly provided https://git.kernel.org/netdev/net/c/986afaf80994 - [net,v8,4/6] net/sched: netem: validate slot configuration https://git.kernel.org/netdev/net/c/01801c359a74 - [net,v8,5/6] net/sched: netem: fix slot delay calculation overflow https://git.kernel.org/netdev/net/c/51e94e1e2fef - [net,v8,6/6] net/sched: netem: check for negative latency and jitter https://git.kernel.org/netdev/net/c/90be9fedb218
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html